We are pleased the Commission has scheduled today's forum on access
to bandwidth. With the rapid emergence of the Internet as the
world's most dynamic information technology, the most vexing problem
concerns the poor choices available for most residential consumers.
Information services that are delivered over today's analog modems
are severely constrained by the limited bandwidth. We are also
faced with a massive public relations effort launched by several
major local exchange carriers (LECs), making a number of alarmist
and often misleading statements about congestion of the public
switched telephone network.
1. Accounts of Network Congestion Are Greatly Exaggerated by
To read current newspaper accounts, one would think the public
switched telephone network is about to collapse due to heavy use
by residential consumers dialing into Internet Service Providers
(ISPs). In fact, there is no evidence that current residential
modem usage is placing greater demands on the public switched
network than non-modem usage. Bell Atlantic and other LECs cite
studies to suggest that residential modem calls are longer than
voice calls, but this is hardly conclusive. One 20 minute modem
call is certainly less of a burden on the network than 100 five
minute voice calls, for example. To the degree that network costs
can be allocated on the basis of usage, one would typically look
at the total number of calls and total minutes of connection.
Because call setup consumes network resources, one long call
is actually cheaper than several shorter calls which sum to the
same total minutes.
What are the data on usage? Let's look at the American Online
Experience. Prior to its new flat rate offering, AOL consumers
reportedly logged on for an average of 14 minutes per day, or
about 7 hours per month. After the flat rate plan was implemented,
usage reportedly jumped to about 32 minutes per day, or about
16 hours per month -- hardly earth shattering numbers. These data
are not that much different from independent studies by Yankelovich,
Nielsen and others, who report average usage of 12 to 22 hours
per month for various online or direct Internet service providers.
The average connect times are likely growing, but they remain
far lower than one might imagine.
There is another way to look at usage. The ISPs have limited
capacity in terms of incoming lines. Before the flat rate, AOL
reportedly was engineered to allow about 3 percent of its consumers
to connect at any one time. Now AOL is expanding its network,
so that it will have about 400,000 incoming lines, for about 8
million customers -- about one line for each 20 customers. This
will permit 5 percent of the AOL customers to be connected at
any one time. Based upon our research, most ISPs have 10 and
20 customers per incoming line, allowing a maximum of 5 to 10
percent of customers to be connected at any one time. This is
a hard constraint on residential consumers. They can't connect
in larger numbers because the ISPs have limited capacity.
How does this compare to capacity on the public switched network?
Bell Atlantic says the residential voice network is built so
that 1 in 7 consumers can connect at any one time -- or roughly
14 percent. Thus, after the AOL expansion of its network, when
it can accommodate 5 percent of its customers, it will be using
at most roughly one third the capacity of the public switched
telephone network. Clearly residential data calls cannot be the
source of congestion of the public switched network.
Of course, residential consumers are calling ISPs, and the ISP
incoming lines may be used intensively. This doesn't have anything
to do with the way the residential consumers call. Regardless
of whether the residential consumers called 5 hours per month
or 150 hours per month, the ISP would have an incentive to purchase
the number of incoming lines that gave customers an acceptable
level of service. Indeed, any business line user makes a similar
For additional discussions of congestion issues, see Fred Goldstein's
November 1996, Maryland ISDN testimony, at
2. We Need To Move Toward An Architecture That Permits Permanent
Digital Residential Connections To A Wide Area Network.
While current levels of residential Internet usage do not place
noticeable demands on the public switched network, one can imagine
a future where everyone, or at least everyone who wanted to, would
have a permanent digital connection to a wide area network. For
this type of connection, one would expect something different
than the current public switched network architecture. There
are likely competing visions of how this can be accomplished:
through a variety of xDSL technologies, cable modems, wireless
connections or other alternatives. This is something we want
to migrate to. One can imagine broad deployment of new technologies
sometime in the next ten years. But what about now?
3. The Best Available Here-And-Now Residential Technology is
If we could use press releases to surf the Internet, everyone
would be using cable modems, ADSL, or satellite communications
to get fabulous last mile bandwidth. Unfortunately, these new
services have been over-promised and under-delivered. ISDN is
the best ubiquitous technology for higher speed connections.
What types of services can be delivered over ISDN? Very high
quality audio, and rough but interesting video. Moreover, ISDN
can provide a very fast launch and disconnect, and multiple ISDN
lines can be bonded together for greater bandwidth. ISDN is something
that can be done right now.
4. High Prices For Residential ISDN Have Stifled Deployment
The problem isn't technology -- it is monopoly pricing by the
LECs. In Japan, residential ISDN is priced about $10 per month
more than a POTS line. In some European markets, there is no
difference between POTS and ISDN rates. Many U.S. businesses
pay no extra for ISDN voice service. But in the United States,
some LECs are charging very high fees for residential ISDN connections.
While interest in the Internet is exploding and there is a great
hunger for more bandwidth, the LECs are killing ISDN as a transition
technology, both through high pricing, and by sending a signal
to consumers and information services developers that the technology
has an uncertain future.
The most annoying features of these high prices are stiff usage
charges. Nynex, Bell Atlantic, U.S. West, PacBell, GTE, and other
LECs have filed tariffs asking for fees of 1 to 3 cents per
minute (per B channel) for making local calls. Meanwhile,
several LECs are providing cost data to PUCs indicating that their
own estimates of direct usage costs are closer to 10 to 12 cents
per hour, or 6 to 17 percent of the per-minute charge billed
to ISDN users. US West and Bell Atlantic offer complex "call
pack" options, that give consumers deep discounts on usage,
but only if they pre-pay for usage they may never need. Under
these plans pre-paid ISDN usage costs $.75 to $.10 per hour (per
channel), depending upon the size of the call pack.. Larger users
typically get the largest and deepest discounts.
Not every state or company has high residential ISDN rates.
While Ameritech charges one of the highest prices in the country
to consumers in Indiana, it offers flat rate ISDN service in Ohio,
Wisconsin, Illinois and Michigan at $26 to $34 per month. Bell
South charges $79 in North Carolina, but only $30 in Tennessee.
U.S. West rates vary from state to state, depending largely on
the views of state regulators. In Utah, the PSC recently ordered
a set of call-pack and flat rate options which ranged from $30,
for 50 hours, to $64, for flat rate service. Roseville, an independent
telephone company in California, offers a voluntary flat rate
tariff of $29. The largest California LEC is PacBell, and it
now charges $24.50 per month, with unmetered usage after 5pm and
on weekends. (See also,
In our view, one of the most interesting telephone companies
in the United States is the Northern Arkansas Telephone Company
(NATCO). NATCO is a surprisingly sophisticated rural LEC, run
by Dr. Steven Sanders, who holds degrees in engineering and physics.
NATCO offers flat rate residential ISDN for $17.90 per month,
roughly $7 more than the cost of POTS. To encourage ISDN usage,
NATCO offers free Internet connectivity to its customers. Dr.
Sanders believes ISDN is potentially very important for rural
communities, who are unlikely to see deployment of cable modems
or ADSL for many years.
5. NTS Costs Of Residential ISDN Are Less Than $5 Per Month
More Than POTS
In the recent FCC proceeding on ISDN and the Subscriber Line
Charge, five LECs reported the ratio of POTS and BRI ISDN costs
for local loop only costs, as well as for all non-traffic sensitive
(NTS) costs. This data was reported in the current NPR, on page
36. As a group, the companies reported local loop costs lower
for ISDN than for POTS. For all NTS costs, the companies say
BRI ISDN costs are 24 percent higher than POTS. If NTS costs
are in the neighborhood of $20 per BRI line, the additional costs
of ISDN would be less than $5 per line. These are averages, and
some companies reported lower figures, of course. For example,
US West said the NTS cost of BRI ISDN was $18.52, only $1.18 higher
than its NTS cost for POTS. (In the Matter of End User Common
Line Charges, FCC Docket No. 95-72, Appendix A , filed June 29,
When competition exists, LECs have priced ISDN more aggressively.
For example, ISDN Centrex is often priced far below available
residential rates. In at least one Bell Atlantic state, the LEC
claimed that usage costs for ISDN Centrex, a service facing competition
from PBX vendors, were less than $1 per month.
6. Regulators Need To Keep One Eye On The Present, And One
Eye On The Future.
Government regulators and policy makers must live in the present
and plan for the future. If new technologies like cable modems
or ADSL are truly seven years from widespread deployment, we must
do something about the present. To put things into perspective,
consider that seven years ago the Internet was barely visible
to most Americans, 2,400 baud modems were fast, Netscape and other
graphical web browsers had not been developed, there was no RealAudio,
no Yahoo, and no intelligent agents to gather information for
consumers. Seven years is a long time for the Internet, and if
we have something much better than analog modems now, we shouldn't
wait seven years before we have faster residential connections.
We also want to move toward a true data network. Vendors of
network facilities now believe they can take incoming calls at
the Central offices and route data calls to ISPs on a more efficient
data network before they consume scarce switch and interoffice
trunk resources. Unfortunately, Bell Atlantic and other LECs
are not pricing this service competitively.
7. Regulators Need To Protect ISPs Against Anticompetitive
Actions By LECs.
Bell Atlantic wants regulators to increase tariffs selectively
on ISPs, so that POTS and ISDN lines will be priced high enough
that the new packet switched services will become more attractive.
If I owned Bell Atlantic, I'd ask for this too. Not only will
this help Bell Atlantic market a new service, but it will raise
costs for unaffiliated ISPs, which compete against Bell Atlantic's
own in-house ISP service.
The Commission should be wary of proposals that would impose new
fees on ISPs which compete with LECs. In California, PacBell
was petitioning the FCC to eliminate the ISPs enhanced service
exemption (ESP), because of the alleged congestion problems that
Internet users are imposing on the public switched network. At
the same time, PacBell was promoting second residential telephone
lines, priced at $11.25 per month, by running promotions that
Order another home phone line today and get 5 months free unlimited
Internet access with Pacific Bell Internet. Act now and we'll
waive the usual $14.95 sign-up fee for Pacific Bell Internet.
Aside from the outrageous hypocrisy of running both this promotion
and a PR battle that asserted Internet users were causing
unbearable network congestion, it demonstrates the dangers facing
independent ISPs. If you ran a California ISP, wouldn't it be
hard to compete with a firm that was giving the service away free
for 5 months?
Indeed, several LECs are running promotions for residential second
POTS lines. These are typically priced below what the company
claims are its average NTS costs. The lines also do not typically
generate any new toll traffic. (People don't get second lines
to make two long distance calls at the same time). Since these
second lines are marketed to modem users, one has to doubt the
LECs complaints about network congestion. It is also reasonable
to assume that the $11.25 per month for a PacBell second line
is actually above PacBell's costs.
8. Possible Incentives For Better Bandwidth Management.
Here are a few thoughts on possible incentives for better bandwidth
A. Flat rate or so called "virtual flat rate" (a service
with a 200 hour usage allowance) tariffs give the LECs an incentive
to solve bandwidth management problems. Metered usage tariffs
take those incentives away, since anything that gets consumers
off the meter reduces LEC revenues. Push for flat rate tariffs,
or tariffs with large usage allowances, such as the 200 hour allowances
proposed for PacBell. If possible, don't meter off-peak usage.
B. Rather that impose new usage fees on ISPs or residential users
of POTS, consider the elimination of usage charges on ISDN and
other higher bandwidth digital services, when calls are routed
away from the circuit switched network to a more efficient packet
switched network. Give ISPs and consumers a positive incentive
to use both higher bandwidth technologies and the packet switched
C. Reward LECs for good deployment of ISDN or other digital services,
punish LECs for poor deployment. I would use a formula that would
compare each LECs penetration of residential ISDN and other digital
services to the average penetration level for all LECs. LECs with
poor penetration should make larger payments to the Universal
Services Fund than LECs with good penetration.
D. Unbundle network elements on a flat rate basis, so that new
entrants can deploy new digital services without high usage charges.
Don't permit LECs to charge per minutes fees for the local loop.
E. Officials in Japan are concerned that the U.S. is far ahead
on the development of the Internet, and they attribute this lead
to the U.S.'s flat rate pricing for local telephones. NTT has
introduced a modest flat rate off-peak calling plan, which works
for two telephone numbers -- one for a bulletin board, and one
for an ISP. It would be interesting to know if prior registration
of telephone numbers would make it easier to take data calls off
the circuit switched network.